Purpose:
To evaluate the effect of sample area size on variability inendothelial cell density (ECD) results. Given that the rateof change in ECD (cells/mm2)=1,000,000/sample area (in µm2),the objective is to determine a recommended minimum sample sizeby comparing results of cell counts performed within largerareas vs. smaller areas.
Methods:
Variable-frame counts were prospectively performed on randomlyselected donor specular images based on approximate sample areasof >60,000µm2, 60,000µm2, 50,000µm2 and40,000µm2 and compared with multiple-field counts basedon sample areas of 20,000-25,000µm2.
Results:
30 donor specular images were analyzed, with ECD ranging fromapproximately 1,700-3,500 cells/mm2. ECD results were groupedinto Class A (40,000-70,000µm2 sample area) and ClassB (20,000-25,000µm2 sample area) to calculate mean densityfor comparison. The mean ECD of Class B was greater than thatof Class A in every instance. Relative percent difference betweenClass A and Class B ranged from 3.7% to 13.2% with a mean RPDof 7.9%.
Conclusions:
Findings indicate cell counts performed in smaller sample sizesproduce significantly higher ECD than counts performed in largersample sizes due to the compounding effects of area and perimeteron the potential for human error. An ideal sample size of >60,000µm2and minimum sample size of 50,000µm2 is recommended foryielding ECD results with greatest statistical accuracy.
Keywords: cornea: endothelium • imaging/image analysis: clinical • image processing